Multi-Step Wind Speed Forecasting Using Signal Decomposing Algorithms, Bat Optimization Algorithm and Least Squares Support Vector Machine

Accurate forecasting of short term wind speed has been widely applied in the disaster early warning of civil engineering. Considering the characteristics of non-stationary and nonlinear of wind speed, the actual wind speed time series need to be decomposed first and then predicted. In this paper, a set of actual wind speed time series of typhoon is decomposed by four signal decomposing algorithms. (e.g., Wavelet Packet Decomposition/Ensemble Empirical Mode Decomposition/Variational Mode Decomposition/Empirical Wavelet Transform) And the features of intrinsic mode functions from these four methods are fully evaluated. Finally, multi-step wind speed forecasting experiment based on least squares support vector machine optimized by bat algorithm is carried out to testify the effectiveness of the signal decomposing algorithm. The results of experiences indicate that the Empirical Wavelet Transform is effective in the wind speed accurate forecasting.